9.5 Update 4b, includes GFS fix?

[ferrus]

Had a recent support call (#03515856), which identified that we were affected by a post 9.5u4 bug.
The bug has resulted in a couple of months of lost backups for us (via broken GFS rotation - rather that underlying data loss), but now it's identified we can work around it.

In the call it was mentioned that the fix was scheduled for update 5, but I noticed in Gostev's last weekly digest that there will now be a 9.5 Update 4b - in June.
Do you know if the fix is scheduled for that, or is there still a subsequent 9.5 update 5 coming later?

[Rick.Vanover]

Hi Ferrus - I recommend checking the release notes of u4b that may include this specific note to GFS logic fix - or if not in the release notes, inquire about this case with support after the release of 4b to see if it includes this fix.

[omegagx]

can you describe the exact conditions that manifest this bug? I'd like to avoid it until then if I can.
Thanks a lot.

Had a recent support call (#03515856), which identified that we were affected by a post 9.5u4 bug.
The bug has resulted in a couple of months of lost backups for us (via broken GFS rotation - rather that underlying data loss), but now it's identified we can work around it.

In the call it was mentioned that the fix was scheduled for update 5, but I noticed in Gostev's last weekly digest that there will now be a 9.5 Update 4b - in June.
Do you know if the fix is scheduled for that, or is there still a subsequent 9.5 update 5 coming later?

[Steve-nIP]

Yeah, it sounds pretty interesting. Could you explain?

[ferrus]

I didn't manage to identify the exact conditions on the bug in our case - as we were time limited - and re-ordering the backups anyway (in the last week, we've actually migrated some of the affected backups onto Veeam Agents).
So there was little point for us to continue testing.

The support engineer did mention a few conditions, where the bug has arisen ...

GFS retention - applied
Read entire point option - applied
Monthly restore points - scheduled on the same day as Weekly restore points

Many of our Backup Copy Jobs fall under these conditions - but only some of them have been affected by the bug, post u4.
There was some questions regarding the scheduling of the jobs making a difference, but it was at this point we ended the call.

Basically, the fix is to schedule Monthly and Weekly backups on different days - which is a pain if you have jobs that require a large backup window.

You can check if the bug exists (only a minority of our BCJ had it), if the GFS restore point you expected to be marked Monthly M (eg. first Saturday of the month), only has the Weekly W tag.
If your GFS schedule is to keep 4 Weekly backups for example - those Monthly restore points will be deleted with the Weekly schedule.

[foggy]

Unfortunately, information provided by support was not entirely correct - the issue is not specific to U4, it existed in previous releases as well (at least in U3), nor the fix for this issue was scheduled for U5. Addressing it could affect too many places, so it is not an easy one.

Aside from re-scheduling monthly and weekly backups on different days, you could also set a backup copy interval to start at 12 AM, since the issue shows up in cases where it starts at non-default time only.

[ferrus]

Thanks foggy. The backup job re-order, that I mentioned we were doing, moved the issue from some jobs - onto others.
I've already implemented the advice from the call - to separate the Weekly and Monthly restore points, and that has worked.

Good to know that there's another way to resolve it. I may try that if it's more efficient.

[morrie]

Hi All,

We use Retention policies with our StoreOnce Dedup appliance and appear to be using the same settings as Ferrus listed belong..
I don't think we are affected by this; The Backup Copy job has both W M and Q. We have mostly W's and separate Q's, but we have M's and Q's together sometimes.

Is there any proper guidance as to this issue (seeing as though it is a known issue); what the particulars are as far as Veeam understand, and what the workarounds are until a proper fix is available. I'm assuming the aim is to fix it?

Andrew Morison
Techno Group.

[JHuston]

Interesting!

I had this exact bug I opened a case for in Nov of 2016 (01967519) for backup copies to our NetApp Altavault . Supports workaround was to schedule the jobs on different days just like you describe. This was later fixed by one of the version releases back then and has stayed working even in 9 U3 which I'm currently on.